[tn] support english text norm 英文tn支持计划

[xingchensong]

NOTE

并不只是简单照搬nemo的规则，在其基础上进行了简化，fst编译时间从460s降低到56s，fst大小从40M降低到4M，同时基本功能依旧

高优

• basic component, like cardinal and word (整数，12345) feat(tn): [cr_id_skip] Support english tn, cardinal and word #203 @xingchensong
• ordinal (序数，1st -> first) [tn] english tn, support ordinal #204 @xingchensong
• date (日期，2024/04/03) [tn] english tn, support date #205 @xingchensong
• decimal (小数，12.01 dollers) [tn] english tn, support decimal #207 @xingchensong
• fraction (分数，1/2) [tn] english tn, support fraction #209 @xingchensong
• time （时间，2:30 a.m.）[tn] english tn, support time #210 @xingchensong
• measure (量词，23kg) [tn] english tn, support measure #211 @xingchensong
• money （钱，$23） [tn] english, support money #212 @xingchensong
• whitelist (白名单) [tn] english tn whitelist for asr and tts #206 [tn] english tn, support whitelist #216 @whiteshirt0429

低优

• telephone (电话号码，010-2345-232323) [tn] english, support telephone #213 @xingchensong
• electronic (URLs, email addresses，[email protected]) [tn] english, support electronic #214 @xingchensong
• roman (罗马数字，IV->four) [tn] tn english, support roman #215

低低优 (暂时可以不实现)

serial (数字和字母，特殊字符组合，c2365-zh)

• range (范围, 10% to 20%) [tn] english tn, support range #233

ref:

• build_tagger(): https://github.com/NVIDIA/NeMo-text-processing/tree/main/nemo_text_processing/text_normalization/en/taggers
• buid_verbalizer(): https://github.com/NVIDIA/NeMo-text-processing/tree/main/nemo_text_processing/text_normalization/en/verbalizers

TODO

• c++ runtime 适配 [runtime] support english tn #219 [runtime] fix english tn #220
• 编译优化，目前编译时间接近十分钟，fst过大，超过40M [tn] simplify tn #221

BUGFIX

• [fix] english tn #224
• [tn] support punct #225
• [fix] remove punc in decimal #226
• [fix] remove punc in measure #227
• [fix] english tn, whitelist exclude punct #228